US8620087B2 - Feature selection device - Google Patents

Feature selection device Download PDF

Info

Publication number
US8620087B2
US8620087B2 US13/056,767 US201013056767A US8620087B2 US 8620087 B2 US8620087 B2 US 8620087B2 US 201013056767 A US201013056767 A US 201013056767A US 8620087 B2 US8620087 B2 US 8620087B2
Authority
US
United States
Prior art keywords
features
image
types
images
feature extraction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active, expires
Application number
US13/056,767
Other languages
English (en)
Other versions
US20110135203A1 (en
Inventor
Kota Iwamoto
Ryoma Oami
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
NEC Corp
Original Assignee
NEC Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by NEC Corp filed Critical NEC Corp
Assigned to NEC CORPORATION reassignment NEC CORPORATION ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: IWAMOTO, KOTA, OAMI, RYOMA
Assigned to NEC CORPORATION reassignment NEC CORPORATION ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: IWAMOTO, KOTA, OAMI, RYOMA
Publication of US20110135203A1 publication Critical patent/US20110135203A1/en
Application granted granted Critical
Publication of US8620087B2 publication Critical patent/US8620087B2/en
Active legal-status Critical Current
Adjusted expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/211—Selection of the most significant subset of features
    • G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation

Definitions

  • the present invention relates to devices for selecting features, and in particular, to devices for selecting features suitable for image signatures for discriminating images (determining identity of images).
  • Image signatures are image features for discriminating images (determining identity). By comparing an image signature extracted from an image with an image signature extracted from another image, an identity scale (in general, referred to as similarity or distance) indicating a degree of the two images being identical can be calculated from a comparison result. Further, by comparing the calculated identity scale with a threshold, it is possible to determine whether or not the two images are identical.
  • an identity scale in general, referred to as similarity or distance
  • the meaning of “two images being identical” includes not only the case where the two images are identical at the level of image signals (pixel values of pixels constituting the images), but also the case where one image is a duplicate image of the other by means of various alteration processes such as conversion of compression format of an image, conversion of size/aspect ratio of an image, adjustment of color tone of an image, various filtering processes (sharpening, smoothing, and the like) applied to an image, local processing (caption superimposition, cutout, and the like) applied to an image, and recapturing of an image.
  • image signatures as it is possible to detect duplication of an image or a moving image which is a collection of images, for example, image signatures are applicable to an illegal copy detection system for images or moving images.
  • An image signature is generally formed of a collection of features. Given that each of the features included in a collection is a dimension, an image signature is composed of feature vectors of multiple dimensions. In particular, a quantization index (quantized value), which is a discrete value, is often used as a feature. Examples of image signatures are described in Non-Patent Document 1, Non-Patent Document 2, and Patent Document 1. In the methods described in those documents, features are extracted for a plurality of local regions of an image, the extracted features are quantized to obtain quantization indexes, and the calculated quantization indexes for the respective local regions constitute quantization index vectors to serve as image signatures.
  • a quantization index quantized value
  • Non-Patent Document 1 and Non-Patent Document 2 an image is divided into blocks. Each of the blocks is used as a local region, and a feature (quantization index) is extracted. Further, in Non-Patent Document 1, luminance distribution patterns within a block are classified into eleven types and are used as quantization indexes. In Non-Patent Document 2 (art described as “Local Edge Representation” in Non-Patent Document 2), a position of center of gravity of an edge point, extracted from a block, is quantized to be used as a quantization index.
  • the method described in Patent Document 1 includes respectively calculating mean luminance values from thirty two pieces of rectangle regions 244 (among them, sixteen pieces of rectangle regions are shown in FIG. 5 ) at predetermined positions in an image 240 , and calculating differences in mean luminance value between rectangle regions forming pairs (the paired rectangle regions are linked to each other with dotted lines 248 in FIG. 5 ), to thereby obtain a difference vector 250 in sixteen dimensions.
  • a composite vector is generated by means of vector transformation, and a quantization index vector in sixteen dimensions, acquired by quantizing the respective dimensions of the composite vector, is used as an image signature.
  • selecting features to be used is important because it determines performance (accuracy of determining identity of images) of the image signature.
  • performance of the image signature can be improved by appropriately selecting the features.
  • each of the features is extracted from a local region determined for each feature (different from each other). As such, in the examples of those documents, performance of the image signature is determined depending on a local area from which the feature is extracted (what kind of local area is set for each feature).
  • Non-Patent Documents 1 and 2 a local region for each of the features is a block formed by regularly dividing an image.
  • a feature selection technique is used in the field of pattern recognition.
  • methods using Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) have been known.
  • LDA Linear Discriminant Analysis
  • an object of the present invention is to provide a feature selection device capable of solving a problem that it is difficult to optimize performance (discrimination accuracy of the identity of images) of image signatures.
  • a feature selection device includes a feature extraction unit that extracts M types of features from each of a plurality of original images and each of a plurality of altered images obtained by applying an alteration process to the plurality of original images; and a feature selection unit that handles original images and altered images of the original images as identical images and handles altered images of the same original image as identical images, while handles other images as different images, and with use of discrimination capability which is a degree of discriminating different images and robustness which is a degree that a value of a feature does not vary due to the alteration process applied to an image as evaluation criteria, evaluates the M types of features extracted from the respective images, and selects a collection of N types of features, the N types being smaller in number than that of the M types, from the M types of features as features for discriminating images.
  • the present invention is configured as described above, the present invention is able to optimize performance (discrimination accuracy of the identity of images) of image signatures formed of a collection of features for discriminating images.
  • FIG. 1 is a block diagram showing a first embodiment of the present invention.
  • FIG. 2 is an illustration showing a method of extracting multi-shape region comparison features.
  • FIG. 3 is an illustration showing exemplary data stored in a feature storing unit.
  • FIG. 4 is a flowchart showing exemplary processes in the first embodiment of the present invention.
  • FIG. 5 is an illustration showing a method of extracting image signatures described in Patent Document 1.
  • FIG. 6 illustrates a feature extraction device according to another embodiment of the invention.
  • a feature extraction device uses a group of images included in an image database to select, from M types of features, N types of features which are smaller in number than that of M types (N ⁇ M) suitable for an image signature, and output information indicating the selected N types of features.
  • features suitable for an image signature means features enabling to improve determination accuracy of the identity of images.
  • a group of N types of features, selected by the feature selection device according to the invention are used as features of respective dimensions in N-dimensional feature vectors (image signature).
  • N-dimensional feature vectors image signatures
  • a method of matching N-dimensional feature vectors that is, as a method of calculating an identity scale, a method of calculation based on comparison between values of identical features (in a corresponding dimension) (for example, calculating the number of dimensions in which feature values (quantization indexes) matches as a similarity or calculating a Hamming distance, a Euclidean distance, a cosine similarity (inner product) or the like).
  • M and N are not necessarily predetermined numerical values (constant numbers) but may be indefinite numbers if they are positive integer numbers satisfying M>2 and N ⁇ M.
  • the feature extraction device includes an image alteration unit 11 , a feature extraction parameter generation unit 12 , a feature extraction unit 13 , a feature selection unit 14 , an original image storing unit 21 , an altered image storing unit 22 , a feature extraction parameter storing unit 23 , and a feature storing unit 24 .
  • the original image storing unit 21 is an image database which stores a plurality of original images in association with image IDs such as numbers for uniquely identifying the respective original images.
  • the original images stored in the original image storing unit 21 are used for selecting features suitable for image signatures by the feature selection device according to the present embodiment.
  • the original images stored in the original image storing unit 21 are supplied to the image alteration unit 11 and the feature extraction unit 13 .
  • the group of original images stored in the original image storing unit 21 is used for selecting features suitable for image signatures, it is desirable to include as many original images as possible (for example, not less than ten thousand images). Further, as the group of original images stored in the original image storing unit 21 is used for selecting features suitable for image signatures, it is desirable that it is a group of images having the same trend (similar trend) to that of an object group of images for which image signatures generated from the features selected by the feature selection device of the present embodiment are used.
  • an object for which image signatures are used is an image or a moving image on the Internet (for example, if image signatures are used for detecting an illegal copy of an image or a moving image on the Internet)
  • the group of original images stored in the original image storing unit 21 should be a group of images taken evenly and uniformly as samples of the images on the Internet.
  • an object for which image signatures are used is limited to a landscape image, for example, it is desirable that the group of original images stored in the original image storing unit 21 should be a group of images taken as samples from various landscape images.
  • the group of original images stored in the original image storing unit 21 should be a group of images taken as samples from various painting images. Further, if various types of images are mixed in an object group of images for which image signatures are used, it is desirable that the group of original images stored in the original image storing unit 21 should include various types of images in the same ratio as that of the object group of images.
  • the group of original images stored in the original image storing unit 21 is a group of images having the same trend (similar trend) to that of an object group of images for which image signatures are used, as it is possible to select more suitable features as image signatures for distinguishing an image with respect to the object group of images, image signatures having higher determination accuracy of the identity of images can be configured.
  • the image alteration unit 11 applies an alteration process to an original image supplied from the original image storing unit 21 to generate an altered image.
  • the generated altered image is stored in the altered image storing unit 22 in association with the original image, which is the source, such that the original image of the generated altered image is clearly indicated.
  • any methods may be used.
  • One exemplary method is to apply an image ID to the original image, and apply a value, in which a branch number for uniquely discriminating each of altered images generated from the original image is linked, to each of the altered images as an image ID of the altered image.
  • Exemplary alteration processes include the following processes. However, these are only given as examples, and the alteration processes are not limited to them. Further, the image alteration unit 11 may perform a combination of these processes (e.g., (A)+(D)).
  • the image alteration unit 11 may perform various types of alteration processes to each of the original images stored in the original storing unit 21 to generate a plurality of types of altered images (this means that the number of altered images may be larger than the number of original images).
  • the alteration process performed by the image alteration unit 11 is the same as (has the same trend as that of) an alteration process applied to an image for which image signatures generated from the features selected by the feature selection device of the present embodiment are used, and an alteration process in which robustness is required in the object for which the image signatures are used.
  • the image alteration unit 11 performs alteration processes of (A), (B), (C), and (D).
  • the image alteration unit 11 performs various types of alteration processes at the same rate as that performed on the object.
  • alteration processes which are the same as (have the same trend as that of) the alteration processes performed on the object for which image signatures are used or the alteration processes required robustness, it is possible to select more robust features as image signatures for discriminating an image with respect to the object, whereby it is possible to configure image signatures having higher determination accuracy of the identity of images.
  • the feature extraction parameter generation unit 12 generates a feature extraction parameter, which is a parameter for extracting a feature from an image, for each of the M types of features.
  • the generated M types of feature extraction parameters are stored in the feature extraction parameter storing unit 23 .
  • M type of parameters generated by a unit identical or similar to the feature extraction parameter generation unit 12 or created manually, are stored in the feature extraction parameter storing unit 23 beforehand.
  • M may take any numerical value if it is larger than N which is the number of features selected as features of an image signature (N ⁇ M), M is desirably several times or tens of times larger than N.
  • N the number of the features selected as features of an image signature
  • M is desirably about 2000 to 5000 or larger.
  • the number M becomes larger (by generating more types of features), a possibility of selecting more appropriate N types of features becomes higher, whereby it is possible to configure an image signature having higher determination accuracy of the identity of images.
  • the M types of features extracted by the M types of feature extraction parameters may be of any types. However, the features are desirably improved so as to be effective to larger types of images. An example thereof will be described with reference to FIG. 2 .
  • FIG. 2 is an illustration showing an extraction method of exemplary features (hereinafter referred to as multi-shape region comparison features) which are improved to be effective with respect to a larger number of types of images.
  • multi-shape region comparison features two extraction regions (a first extraction region and a second extraction region) within an image are set beforehand to extract features for each dimension of a feature vector.
  • a significant difference between the multi-shape region comparison feature and the feature described in Patent Document 1 is that shapes of extraction regions have variations.
  • the mean luminance value of the first extraction region and the mean luminance value of the second extraction region are compared, and if the mean luminance value of the first extraction region is larger, the quantization index is set to be +1, while the quantization index is set to be ⁇ 1 in other cases.
  • the quantization index Qn in the dimension n can be calculated from the following formula.
  • the feature extraction parameter corresponding to the feature is information indicating the first extraction region and the second extraction region of each feature.
  • a collection of pixel coordinate values of a first extraction region and a collection of pixel coordinate values of a second extraction region in a normalized image size may be feature extraction parameters.
  • the extraction regions may be represented by fewer parameters. For example, if the shape of an extraction region is a quadrangle, the coordinates of the four corners of the quadrangle may be used as a feature extraction parameter, while if the shape of an extraction region is a circle, the coordinates of the center of the circle and the radius value may be used as a feature extraction parameter. Further, if a threshold th is different for respective types of features, the threshold th may be included in the feature extraction parameter.
  • the feature extraction parameter generation unit 12 may automatically generate feature extraction parameters for M types of multi-shape region comparison features according to pseudo random numbers, for example. For example, it is possible to generate a random number string from seeds of pseudo random numbers, and based on the generated random numbers, automatically generate shapes of extraction regions or thresholds th. For example, if the shape of an extraction region is a quadrangle, it is possible to automatically generate coordinates of the four corners of the quadrangle based on the generated random numbers.
  • the M types of features have an identical characteristic of a mean luminance value.
  • the M types of features are not limited to features having the same characteristics.
  • Features having different characteristics such as color information, frequency information, and edge information may be mixed in the M types of features.
  • feature extraction parameters may be generated in which part of the M types of features is multi-shape region comparison features, part thereof is features based on color information, part thereof is features based on frequency information, and part thereof is features based on edge information.
  • the feature extraction unit 13 receives the original image stored in the original image storing unit 21 and the altered image stored in the altered image storing unit 22 , and extracts M types of features from the original image and the altered image respectively, according to the feature extraction parameters providing methods of extracting the M types of features stored in the feature extraction parameter storing unit 23 .
  • the feature extraction unit 13 has two extraction sections, which are a feature extraction unit 131 for extracting the M types of features from the original image, and a feature extraction unit 132 for extracting the M types of features from the altered image.
  • the two feature extraction units 131 and 132 may be operated in parallel or in order.
  • the M types of features extracted from the original image and the M types of features extracted from the altered image are stored in the feature storing unit 24 .
  • the feature storing unit 24 stores the M types of features extracted from the original image and the M types of features extracted from the altered image in association with each other. For associating them with each other, any methods may be used. For example, if the image ID of the altered image is one in which a branch number is added to the image ID of the original image, applying the image ID of the original image to the group of M types of features extracted from the original image, and applying the image ID of the altered image to the group of M types of features extracted from the altered image of the original image. Thereby, by identifying the image ID, association between the M types of features extracted from the original image and the M types of features extracted from the altered image of the original image can be recognized.
  • FIG. 3 shows an example of data stored in the feature storing unit 24 .
  • the feature storing unit 24 of this example stores groups of features 24 - 1 to 24 - x , corresponding one to one to the original images stored in the original image storing unit 21 .
  • One group of features that is, a group of features 24 - 1 for example includes original image feature data 241 consisting of the image ID of the original image and M types of features extracted from the original image, and a plurality of altered image feature data 241 - 1 to 241 - y each consisting of the image ID of an altered image of the original image and M types of features extracted from the altered image.
  • images belonging to the same group of features that is, one original image and its altered images
  • images belonging to different groups of features are handled as different images.
  • the feature selection unit 14 uses the values of the M types of features extracted from the original image stored in the feature storing unit 24 and the values of the M types of features extracted from the altered images to select a collection of N types of features such that the sum of the discrimination capability which is a degree of discriminating different images, held by the collection of features, and the robustness which is a degree that the values of the features do not vary due to alteration processes performed on the images, held by the collection of features, becomes larger.
  • the function of the feature selection unit 14 will be described in detail.
  • the value of a feature extracted from the group of original images is represented as a probability variable X n
  • the value of a feature extracted from the group of altered images is represented as a probability variable X′ n .
  • the feature can be regarded as a probability variable which takes any value (quantization index) of ⁇ +1, 0, ⁇ 1 ⁇ .
  • a larger value of D(S N ) shows higher discrimination capability.
  • a larger value of R(S N , S′ N ) shows higher robustness.
  • the feature selection unit 14 selects collections of N types of features such that the value of E(S N , S′ N ) given by the evaluation formula 2 becomes larger. For example, collections of N types of features which allow the value of E(S N , S′ N ) to be maximum may be selected (in a lump). Alternatively, features may be selected (added) sequentially such that the value of E(S N , S′ N ) becomes larger to thereby select collections of features.
  • a specific example of a method of calculating discrimination capability D(S N ) and robustness R(S N , S′ N ), in the case where values of features are discrete values (quantization index), will be described.
  • ⁇ n ⁇ +1, 0, ⁇ 1 ⁇ .
  • the discrimination capability held by a collection of features can be larger as the information entropy of respective features is larger, for example.
  • the appearance probability of values (probability variables X n ) taken by the respective features becomes close to uniform, so that redundancy is decreased and the discrimination capability is increased.
  • appearance probabilities of values (probability variables Xn) which may be taken by the respective features are biased toward particular values, redundancy is increased, and as the information entropy is decreased, the discrimination capability is decreased.
  • the discrimination capability D(S N ) held by the collection of features can be calculated as the sum of information entropies of the respective features.
  • H ( X n ) of the probability variable X n of a feature n is calculated from the following formula.
  • H ( X n ) ⁇ AA p ( x n )log p ( x n ) [Formula 3]
  • the feature n is a multi-shape region comparison feature described above
  • p(+1) Pr(X n +1)
  • the discrimination capability D(S N ) held by a collection of features can be calculated as the sum of information entropies H(X n ) of the respective features, for example, from the following formula.
  • the discrimination capability of a collection of features becomes larger if correlation between the respective features is lower and probability independency is higher, because redundancy is lower. If features having large correlation between them are collected, redundancy becomes higher, so that the discrimination capability for discriminating images will not be improved.
  • a scale indicating correlation between the respective features mutual information amount can be used.
  • (X n ;X k ) between the probability variable X n of a feature n and the probability variable X k of a feature k is calculated from the following formula:
  • ( X n ;X k ) ⁇ BB ⁇ CC p ( x n ,x k )log [ p ( x n ,x k )/ ⁇ p ( x n ) p ( x k ) ⁇ ] (Formula 5)
  • p(x n , x k ) is calculated from values of features of a group of original images to be supplied. For example, if the features are multi-shape region comparison features described above, the following equations are established.
  • the simultaneous probability may be calculated from the appearance frequency of combinations of the feature n and the feature k of the group of original images.
  • the discrimination capability D(S N ) of the collection of features can be calculated by the following formula, based on the sum of all combinations of the features.
  • the reason of the sum of the mutual information amount being negative (“ ⁇ ” is attached) is that the value of the discrimination capability D(S N ) is larger as the sum of the mutual information amount is smaller.
  • discrimination capability D(S N ) of a collection of features may be calculated according to the following formula using the sums of Formula 4 and Formula 6.
  • the discrimination capability D(S N ) may be calculated according to the following formula.
  • the discrimination capability may be calculated from the value of a feature (probability variable X′ n ) of a group of altered images, rather than the value of a feature (probability variable X n ) of the group of original images, or may be calculated from a combination thereof.
  • the feature is a multi-shape region comparison feature described above
  • the robustness R(S N , S′ N ) of a collection of features may be calculated based on a conditional entropy H(X n
  • X′ n ) of a feature n indicates the ambiguity of the probability variable Xn (value of the feature in the group of original images) which remains when the probability variable X′n (value of the feature in the group of altered images) has been known, if the probability that the value of feature does not vary before and after the alteration is high (probability that the features coincide before and after the alteration is high), that is, if the robustness if high, the conditional entropy H(X n
  • X′ n ) The conditional entropy H(X n
  • X′ n ) ⁇ EE ⁇ FF p ( x n ,x′ n )log p ( x n
  • x′ n ) is a conditional probability, which may be calculated from the value of a feature in the group of original images to be supplied and the value of the corresponding feature in the group of altered images. For example, if the feature is a multi-shape region comparison feature described above, the following expressions are established.
  • the robustness R(S N , S′ N ) held by a collection of features may also be calculated according to the following formula, based on the sum of conditional entropies H(X n
  • X′ n ) being negative (“ ⁇ ” is attached) is that the value of the robustness R(S N , S′ N ) is larger as the sum of the conditional entropies H(X n
  • the sum E(S N , S′ N ) of the discrimination capability and the robustness may be calculated by combining any one of Formula 4, Formula 6, Formula 7, and Formula 8 as a method of calculating the discrimination capability D(S N ) and any one of Formula 9 and Formula 11 as a method of calculating the robustness R(S N , S′ N ).
  • the sum of the discrimination capability D(S N ) and the robustness R(S N , S′ N ) may be calculated according to the following formula using an appropriate weighting coefficient ⁇ .
  • E ( S N ,S′ N ) ⁇ D ( S N )+(1 ⁇ ) R ( S N ,S′ N ) (Formula 12)
  • the feature selection unit 14 selects the N types of features such that the value of E(S N , S′ N ) by the evaluation formula 2 of the sum of the discrimination capability and the robustness described above becomes larger.
  • the feature selection unit 14 may select a collection of the N types of features in which the value of E(S N , S′ N ) becomes maximum in a lump.
  • a method of selecting a collection of features by sequentially (incrementally) adding features such that the value of E(S N , S′ N ) becomes large is provided as a practicable method.
  • This method is to be called an incremental method.
  • the N types of features are selected by increasing the collections of features one by one.
  • features are added one by one in the following manner.
  • features which allow the value of the evaluation formula 2 of the sum of the discrimination capability and the robustness becomes largest are sequentially selected, and the features are added.
  • a feature (probability variable) Xn which allows the value of E(S N , S′ N ) by the evaluation formula 2 of the sum of the discrimination capability and the robustness becomes maximum, is added.
  • a feature (probability variable) Xn which allows a difference between the value of the evaluation formula 2 of the sum of the discrimination capability and the robustness of the collection of feature before the feature is added and the value of the evaluation formula 2 of the sum of the discrimination capability and the robustness of the collection of features after the feature is added to be maximum, is added.
  • a feature (probability variable) X n which allows the following evaluation formula E ( S N ,S′ N ) ⁇ E ( S N-1 ,S′ N-1 ) (Formula 14) to be maximum, is added.
  • the image alteration unit 11 performs a predetermined type of alteration process on each of the original images read from the original image storing unit 21 to generate an altered image, and stores it in the altered image storing unit 22 (S 101 ).
  • the feature extraction parameter generation unit 12 generates, for each of the M types of features, a feature extraction parameter which is a parameter for extracting a feature from the image, and stores it in the feature extraction parameter storing unit 23 (S 102 ).
  • the feature extraction section 131 of the feature extraction unit 13 extracts the M types of features from each of the original images in the original image storing unit 21 in accordance with the extraction parameters for the M types of features, and stores them in the feature storing unit 24 (S 103 ). Further, the feature extraction section 132 of the feature extraction unit 13 extracts the M types of features from each of the altered images in the altered image storing unit 22 in accordance with the extraction parameters for the M types of features, and stores them in the feature storing unit 24 (S 104 ).
  • the feature selection unit 14 receives the M types of features of the original images and the altered images stored in the feature storing unit 24 , and handles them such that an original image and its altered image, and altered images of the same original image are recognized as identical images to each other while other images are different from each other, and selects N types of features from the M types of features, with the discrimination capability which is a degree of discriminating different images and the robustness which is a degree that the value of a feature does not vary due to an alteration process applied to an image being evaluation criteria, and outputs them (S 105 ). Step S 105 will be described in detail below, using the incremental method as an example.
  • the feature selection unit 14 determines an additional type of feature (S 107 ). Specifically, the feature selection unit 14 selects a type of feature which allows a difference between the value of an evaluation formula of the sum of the discrimination capability and robustness of a collection of features before the feature is added and the value of the evaluation formula of the sum of the discrimination capability and robustness of a collection of features after the feature is added (that is, a value of Formula 14) to be maximum, and the selected type of feature is determined to be a feature which is to be added next.
  • the feature selection unit 14 judges whether or not the N types of features are determined (S 108 ), and if the N types of features have not been determined, the feature selection unit 14 returns to step S 107 and continue to determine the remaining types of features. On the other hand, if the N types of features have been determined, the feature selection unit 14 outputs the determined N types of features to a storing unit not shown in FIG. 1 , for example.
  • features are selected such that, with use of values of features extracted from groups of images before and after alteration, the sum of the discrimination capability which is a degree of discriminating different images and robustness which is a degree that the value of a feature does not vary due to an alteration process applied to an image becomes large.
  • the present invention is not limited to this embodiment. It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the scope of the present invention. Further, the feature extraction device of the present invention is adapted such that the functions thereof can be realized by computers 600 and programs, as well as hardware.
  • Such a program is provided in the form of being written on a computer readable recording medium such as a magnetic disk, a semiconductor memory, or the like, is read by a computer when the computer is started for example, and controls operation of the computer, to thereby allow the computer to function as the image alteration unit 11 , the feature extraction parameter generation unit 12 , the feature extraction unit 13 , the feature selection unit 14 , and the like of the above-described embodiment.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)
US13/056,767 2009-01-29 2010-01-19 Feature selection device Active 2030-02-04 US8620087B2 (en)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
JP2009-017806 2009-01-29
JP2009017806 2009-01-29
PCT/JP2010/000246 WO2010087124A1 (fr) 2009-01-29 2010-01-19 Dispositif de sélection de quantité caractéristique

Publications (2)

Publication Number Publication Date
US20110135203A1 US20110135203A1 (en) 2011-06-09
US8620087B2 true US8620087B2 (en) 2013-12-31

Family

ID=42395390

Family Applications (1)

Application Number Title Priority Date Filing Date
US13/056,767 Active 2030-02-04 US8620087B2 (en) 2009-01-29 2010-01-19 Feature selection device

Country Status (6)

Country Link
US (1) US8620087B2 (fr)
EP (1) EP2333718B1 (fr)
JP (1) JP4766197B2 (fr)
KR (1) KR101404401B1 (fr)
CN (1) CN102301395B (fr)
WO (1) WO2010087124A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120014609A1 (en) * 2009-03-13 2012-01-19 Nec Corporation Image signature extraction device
US20160042517A1 (en) * 2014-08-08 2016-02-11 Robotic Vision Technologies, LLC Detection and tracking of item features

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2333718B1 (fr) * 2009-01-29 2013-08-28 Nec Corporation Dispositif de sélection de quantité caractéristique
US8891878B2 (en) * 2012-06-15 2014-11-18 Mitsubishi Electric Research Laboratories, Inc. Method for representing images using quantized embeddings of scale-invariant image features
CN103218798A (zh) * 2012-01-19 2013-07-24 索尼公司 图像处理设备和方法
US9053359B2 (en) * 2012-06-07 2015-06-09 Konica Minolta Laboratory U.S.A., Inc. Method and system for document authentication using Krawtchouk decomposition of image patches for image comparison
JP6057786B2 (ja) * 2013-03-13 2017-01-11 ヤフー株式会社 時系列データ解析装置、時系列データ解析方法、およびプログラム
US9418283B1 (en) * 2014-08-20 2016-08-16 Amazon Technologies, Inc. Image processing using multiple aspect ratios
US9576196B1 (en) 2014-08-20 2017-02-21 Amazon Technologies, Inc. Leveraging image context for improved glyph classification
JP6048688B2 (ja) * 2014-11-26 2016-12-21 横河電機株式会社 イベント解析装置、イベント解析方法およびコンピュータプログラム
CN104462481A (zh) * 2014-12-18 2015-03-25 浪潮(北京)电子信息产业有限公司 一种基于颜色和形状的综合图像检索方法
WO2017056312A1 (fr) * 2015-10-02 2017-04-06 富士通株式会社 Programme de traitement d'image et dispositif de traitement d'image
CN107045503B (zh) 2016-02-05 2019-03-05 华为技术有限公司 一种特征集确定的方法及装置
KR101986361B1 (ko) * 2016-09-23 2019-06-07 주식회사 모션크루 디지털 동영상 특징값 추출 시스템 및 방법, 및 상기 특징값을 이용한 디지털 동영상 유사도 판단 시스템 및 방법
CN108052500B (zh) * 2017-12-13 2021-06-22 北京数洋智慧科技有限公司 一种基于语义分析的文本关键信息提取方法及装置
DE102019204318A1 (de) * 2019-03-28 2020-10-01 Conti Temic Microelectronic Gmbh Automatische Erkennung und Klassifizierung von Adversarial Attacks
CN111860894B (zh) * 2020-07-29 2024-01-09 宁波大学 斜拉桥病害属性选择方法

Citations (47)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH08500471A (ja) 1992-04-30 1996-01-16 セリディアン コーポレイション 放送セグメントを認識するための方法とシステム
US6016354A (en) * 1997-10-23 2000-01-18 Hewlett-Packard Company Apparatus and a method for reducing red-eye in a digital image
US6157741A (en) * 1997-02-27 2000-12-05 Kabushiki Kaisha Toshiba Image processing apparatus and image processing method
US6292583B1 (en) * 1997-09-30 2001-09-18 Advantest Corporation Image information processing apparatus
US20010052928A1 (en) * 2000-05-22 2001-12-20 Kazuyuki Imagawa Image communication terminal
US20020015523A1 (en) * 2000-05-12 2002-02-07 Seiji Okada Image signal processing device and image signal processing method
JP2002142094A (ja) 2000-10-31 2002-05-17 Toshiba Corp 電子透かし埋込装置、電子透かし検出装置、電子透かし埋込方法、電子透かし検出方法及び記録媒体
US20020138492A1 (en) * 2001-03-07 2002-09-26 David Kil Data mining application with improved data mining algorithm selection
US20020164070A1 (en) * 2001-03-14 2002-11-07 Kuhner Mark B. Automatic algorithm generation
US20020178368A1 (en) * 2001-05-24 2002-11-28 Peng Yin Semi-fragile watermarking system for MPEG video authentication
US6606620B1 (en) * 2000-07-24 2003-08-12 International Business Machines Corporation Method and system for classifying semi-structured documents
US20030161536A1 (en) * 2002-02-27 2003-08-28 Canon Kabushiki Kaisha Information processing apparatus, information processing system, information processing method, storage medium and program
US20030202660A1 (en) * 2002-04-29 2003-10-30 The Boeing Company Dynamic wavelet feature-based watermark
US20030215119A1 (en) * 2002-05-15 2003-11-20 Renuka Uppaluri Computer aided diagnosis from multiple energy images
US20050021317A1 (en) * 2003-07-03 2005-01-27 Fuliang Weng Fast feature selection method and system for maximum entropy modeling
US20050169516A1 (en) * 2002-02-27 2005-08-04 Kenji Okajima Pattern feature selection method, classification method, judgment method, program, and device
US6940999B2 (en) * 2000-06-13 2005-09-06 American Gnc Corp. Method for target detection and identification by using proximity pixel information
US20050276454A1 (en) * 2004-06-14 2005-12-15 Rodney Beatson System and methods for transforming biometric image data to a consistent angle of inclination
US20060093221A1 (en) * 1999-12-10 2006-05-04 Nec Corporation Image retrieval device, image retrieval method and storage medium storing similar-image retrieval program
US20060117183A1 (en) * 2004-11-29 2006-06-01 Yasuo Hatano Digital image data authenticity assuring method, and digital image data disclosure system
WO2006129551A1 (fr) 2005-05-31 2006-12-07 Nec Corporation Procede, systeme et programme de collationnement de modeles
US20070071282A1 (en) * 2005-09-09 2007-03-29 Kabushiki Kaisha Toshiba Digital watermark embedding apparatus and method, and digital watermark detection apparatus and method
US20070100624A1 (en) * 2005-11-03 2007-05-03 Fuliang Weng Unified treatment of data-sparseness and data-overfitting in maximum entropy modeling
WO2007119204A2 (fr) * 2006-04-13 2007-10-25 Medicad S.R.L. Procede de traitement des images biomédicales
US20070253592A1 (en) * 2003-09-26 2007-11-01 Qibin Sun Method and System for Protecting and Authenticating a Digital Image
US20080025638A1 (en) * 2006-07-31 2008-01-31 Eastman Kodak Company Image fusion for radiation therapy
US7346210B2 (en) * 2001-12-28 2008-03-18 Nikon Corporation Image processing device and image processing program for determining similarity factors of pixels
US7356190B2 (en) * 2002-07-02 2008-04-08 Canon Kabushiki Kaisha Image area extraction method, image reconstruction method using the extraction result and apparatus thereof
US20080089552A1 (en) * 2005-08-04 2008-04-17 Nippon Telegraph And Telephone Corporation Digital Watermark Padding Method, Digital Watermark Padding Device, Digital Watermark Detecting Method, Digital Watermark Detecting Device, And Program
US7394940B2 (en) * 2004-11-19 2008-07-01 International Business Machines Corporation Digital video media duplication or transmission quality measurement
JP2008158776A (ja) 2006-12-22 2008-07-10 Canon Inc 特徴検出方法及び装置、プログラム、記憶媒体
US20080279416A1 (en) * 2007-05-11 2008-11-13 Motorola, Inc. Print matching method and system using phase correlation
US7454064B2 (en) * 2000-01-11 2008-11-18 Minolta Co., Ltd. Coder, coding method, program, and image forming apparatus for improving image data compression ratio
US20090232399A1 (en) * 2008-03-11 2009-09-17 Tomokazu Kawahara Apparatus and method for pattern recognition
US20090257656A1 (en) * 2008-04-14 2009-10-15 Yun-Qing Shi Detecting Double JPEG Compression in Images
US20090274364A1 (en) * 2008-05-01 2009-11-05 Yahoo! Inc. Apparatus and methods for detecting adult videos
US20090296989A1 (en) * 2008-06-03 2009-12-03 Siemens Corporate Research, Inc. Method for Automatic Detection and Tracking of Multiple Objects
US7680357B2 (en) * 2003-09-09 2010-03-16 Fujifilm Corporation Method and apparatus for detecting positions of center points of circular patterns
US20100177969A1 (en) * 2009-01-13 2010-07-15 Futurewei Technologies, Inc. Method and System for Image Processing to Classify an Object in an Image
US20110026788A1 (en) * 2008-04-02 2011-02-03 Matthias Elter Method and device for the segmentation of a lesion
US20110135203A1 (en) * 2009-01-29 2011-06-09 Nec Corporation Feature selection device
US20110170784A1 (en) * 2008-06-10 2011-07-14 Tokyo Institute Of Technology Image registration processing apparatus, region expansion processing apparatus, and image quality improvement processing apparatus
US20110199499A1 (en) * 2008-10-14 2011-08-18 Hiroto Tomita Face recognition apparatus and face recognition method
US20110241991A1 (en) * 2009-10-07 2011-10-06 Yasunobu Ogura Tracking object selection apparatus, method, program and circuit
US20110304730A1 (en) * 2010-06-09 2011-12-15 Hon Hai Precision Industry Co., Ltd. Pan, tilt, and zoom camera and method for aiming ptz camera
US20120082370A1 (en) * 2009-04-22 2012-04-05 Nec Corporation Matching device, matching method and matching program
US20120283574A1 (en) * 2011-05-06 2012-11-08 Park Sun Young Diagnosis Support System Providing Guidance to a User by Automated Retrieval of Similar Cancer Images with User Feedback

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE69941499D1 (de) * 1998-10-09 2009-11-12 Sony Corp Vorrichtungen und Verfahren zum Lernen und Anwenden eines Abstand-Transition-Modelles
EP1050848A4 (fr) * 1998-11-25 2003-07-30 Sony Corp Dispositif et procede de traitement d'image, et support enregistre lisible par ordinateur
JP4694535B2 (ja) 2007-07-11 2011-06-08 銀河フーズ株式会社 切れ目入りソーセージの製造方法

Patent Citations (59)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH08500471A (ja) 1992-04-30 1996-01-16 セリディアン コーポレイション 放送セグメントを認識するための方法とシステム
US6157741A (en) * 1997-02-27 2000-12-05 Kabushiki Kaisha Toshiba Image processing apparatus and image processing method
US6292583B1 (en) * 1997-09-30 2001-09-18 Advantest Corporation Image information processing apparatus
US6016354A (en) * 1997-10-23 2000-01-18 Hewlett-Packard Company Apparatus and a method for reducing red-eye in a digital image
US20060093221A1 (en) * 1999-12-10 2006-05-04 Nec Corporation Image retrieval device, image retrieval method and storage medium storing similar-image retrieval program
US7454064B2 (en) * 2000-01-11 2008-11-18 Minolta Co., Ltd. Coder, coding method, program, and image forming apparatus for improving image data compression ratio
US20020015523A1 (en) * 2000-05-12 2002-02-07 Seiji Okada Image signal processing device and image signal processing method
US20010052928A1 (en) * 2000-05-22 2001-12-20 Kazuyuki Imagawa Image communication terminal
US6940999B2 (en) * 2000-06-13 2005-09-06 American Gnc Corp. Method for target detection and identification by using proximity pixel information
US6606620B1 (en) * 2000-07-24 2003-08-12 International Business Machines Corporation Method and system for classifying semi-structured documents
US20060120561A1 (en) * 2000-10-31 2006-06-08 Hirofumi Muratani Digital watermark embedding apparatus, digital watermark detecting apparatus, digital watermark embedding method, digital watermark detecting method and computer program product
US20020071593A1 (en) * 2000-10-31 2002-06-13 Hirofumi Muratani Digital watermark embedding apparatus, digital watermark detecting apparatus, digital watermark embedding method, digital watermark detecting method and computer program product
JP2002142094A (ja) 2000-10-31 2002-05-17 Toshiba Corp 電子透かし埋込装置、電子透かし検出装置、電子透かし埋込方法、電子透かし検出方法及び記録媒体
US20020138492A1 (en) * 2001-03-07 2002-09-26 David Kil Data mining application with improved data mining algorithm selection
US20020164070A1 (en) * 2001-03-14 2002-11-07 Kuhner Mark B. Automatic algorithm generation
US20020178368A1 (en) * 2001-05-24 2002-11-28 Peng Yin Semi-fragile watermarking system for MPEG video authentication
US7346210B2 (en) * 2001-12-28 2008-03-18 Nikon Corporation Image processing device and image processing program for determining similarity factors of pixels
US20030161536A1 (en) * 2002-02-27 2003-08-28 Canon Kabushiki Kaisha Information processing apparatus, information processing system, information processing method, storage medium and program
US20050169516A1 (en) * 2002-02-27 2005-08-04 Kenji Okajima Pattern feature selection method, classification method, judgment method, program, and device
US20070063884A1 (en) * 2002-02-27 2007-03-22 Canon Kabushiki Kaisha Information processing apparatus, information processing system, information processing method, storage medium and program
US20030202660A1 (en) * 2002-04-29 2003-10-30 The Boeing Company Dynamic wavelet feature-based watermark
US7366909B2 (en) * 2002-04-29 2008-04-29 The Boeing Company Dynamic wavelet feature-based watermark
US20030215119A1 (en) * 2002-05-15 2003-11-20 Renuka Uppaluri Computer aided diagnosis from multiple energy images
US7542615B2 (en) * 2002-07-02 2009-06-02 Canon Kabushiki Kaisha Image area extraction method, image reconstruction method using the extraction result and apparatus thereof
US7356190B2 (en) * 2002-07-02 2008-04-08 Canon Kabushiki Kaisha Image area extraction method, image reconstruction method using the extraction result and apparatus thereof
US20050021317A1 (en) * 2003-07-03 2005-01-27 Fuliang Weng Fast feature selection method and system for maximum entropy modeling
US7680357B2 (en) * 2003-09-09 2010-03-16 Fujifilm Corporation Method and apparatus for detecting positions of center points of circular patterns
US20070253592A1 (en) * 2003-09-26 2007-11-01 Qibin Sun Method and System for Protecting and Authenticating a Digital Image
US20050276454A1 (en) * 2004-06-14 2005-12-15 Rodney Beatson System and methods for transforming biometric image data to a consistent angle of inclination
US7522776B2 (en) * 2004-11-19 2009-04-21 International Business Machines Corporation Digital video media duplication or transmission quality measurement
US7394940B2 (en) * 2004-11-19 2008-07-01 International Business Machines Corporation Digital video media duplication or transmission quality measurement
US20060117183A1 (en) * 2004-11-29 2006-06-01 Yasuo Hatano Digital image data authenticity assuring method, and digital image data disclosure system
WO2006129551A1 (fr) 2005-05-31 2006-12-07 Nec Corporation Procede, systeme et programme de collationnement de modeles
JP2006338092A (ja) 2005-05-31 2006-12-14 Nec Corp パタン照合方法、パタン照合システム及びパタン照合プログラム
US20090087036A1 (en) * 2005-05-31 2009-04-02 Nec Corporation Pattern Matching Method, Pattern Matching System, and Pattern Matching Program
US20080089552A1 (en) * 2005-08-04 2008-04-17 Nippon Telegraph And Telephone Corporation Digital Watermark Padding Method, Digital Watermark Padding Device, Digital Watermark Detecting Method, Digital Watermark Detecting Device, And Program
US20070071282A1 (en) * 2005-09-09 2007-03-29 Kabushiki Kaisha Toshiba Digital watermark embedding apparatus and method, and digital watermark detection apparatus and method
US20070100624A1 (en) * 2005-11-03 2007-05-03 Fuliang Weng Unified treatment of data-sparseness and data-overfitting in maximum entropy modeling
WO2007119204A2 (fr) * 2006-04-13 2007-10-25 Medicad S.R.L. Procede de traitement des images biomédicales
US20080118137A1 (en) * 2006-07-31 2008-05-22 Shoupu Chen Method and system of image fusion for radiation therapy
US20080025638A1 (en) * 2006-07-31 2008-01-31 Eastman Kodak Company Image fusion for radiation therapy
US7881516B2 (en) * 2006-07-31 2011-02-01 Carestream Health, Inc. Method and system of image fusion for radiation therapy
US7848592B2 (en) * 2006-07-31 2010-12-07 Carestream Health, Inc. Image fusion for radiation therapy
US20080212098A1 (en) * 2006-12-22 2008-09-04 Canon Kabushiki Kaisha Method and apparatus for detecting feature
JP2008158776A (ja) 2006-12-22 2008-07-10 Canon Inc 特徴検出方法及び装置、プログラム、記憶媒体
US20080279416A1 (en) * 2007-05-11 2008-11-13 Motorola, Inc. Print matching method and system using phase correlation
US20090232399A1 (en) * 2008-03-11 2009-09-17 Tomokazu Kawahara Apparatus and method for pattern recognition
US20110026788A1 (en) * 2008-04-02 2011-02-03 Matthias Elter Method and device for the segmentation of a lesion
US20090257656A1 (en) * 2008-04-14 2009-10-15 Yun-Qing Shi Detecting Double JPEG Compression in Images
US20090274364A1 (en) * 2008-05-01 2009-11-05 Yahoo! Inc. Apparatus and methods for detecting adult videos
US20090296989A1 (en) * 2008-06-03 2009-12-03 Siemens Corporate Research, Inc. Method for Automatic Detection and Tracking of Multiple Objects
US20110170784A1 (en) * 2008-06-10 2011-07-14 Tokyo Institute Of Technology Image registration processing apparatus, region expansion processing apparatus, and image quality improvement processing apparatus
US20110199499A1 (en) * 2008-10-14 2011-08-18 Hiroto Tomita Face recognition apparatus and face recognition method
US20100177969A1 (en) * 2009-01-13 2010-07-15 Futurewei Technologies, Inc. Method and System for Image Processing to Classify an Object in an Image
US20110135203A1 (en) * 2009-01-29 2011-06-09 Nec Corporation Feature selection device
US20120082370A1 (en) * 2009-04-22 2012-04-05 Nec Corporation Matching device, matching method and matching program
US20110241991A1 (en) * 2009-10-07 2011-10-06 Yasunobu Ogura Tracking object selection apparatus, method, program and circuit
US20110304730A1 (en) * 2010-06-09 2011-12-15 Hon Hai Precision Industry Co., Ltd. Pan, tilt, and zoom camera and method for aiming ptz camera
US20120283574A1 (en) * 2011-05-06 2012-11-08 Park Sun Young Diagnosis Support System Providing Guidance to a User by Automated Retrieval of Similar Cancer Images with User Feedback

Non-Patent Citations (12)

* Cited by examiner, † Cited by third party
Title
A. Hampapur, et al., "Comparison of Distance Measures for Video Copy Detection," IEEE International Conference on Multimedia and Expo, 2001, pp. 944-947.
Chen et al. "Quantization Index Modulation: A Class of Provably Good Methods for Digital Watermarking and Information Embedding" ISIT (2000) p. 1-1. *
Choksuriwong et al. "Comparative Study of Global Invariant Descriptors for Object Recognition." Journal of Electronic Imaging, vol. 27, No. 2, pp. 023015-1 through 023015-10, Jan. 1, 2008.
Extended European Search Report dated Oct. 12, 2011 from the European Patent Office in counter application No. 10735596.8.
Fergus et al. "Object Class Recognition by Unsupervised Scale Invariant Learning" Computer Vision and Pattern Recg. (2003) IEEE Computer Society Conference. vol. 2, pp. 1-8. *
Houmansadr et al. "Robust Content Based Video Watermarking Exploiting Motion Entropy Masking Effect" IEEE (Oct. 2004) pp. 1-4. *
Iwamoto et al. "Detection of Wipes and Digital Video Effects Based on a Pattern Independent Model of Image Boundary Line Characteristics" IEEE (2007) pp. 1-4. *
Iwamoto et al. "Image Signature to Caption Superimposition for Video Sequence Identification" IEEE (2006) pp. 1-4. *
K. Iwamoto, et al., "Image Signature Robust to Caption Superimposition for Video Sequence Indentification," IEEE, ICIP 2006, 2006, pp. 3185-3188.
Law-To et al. "Video Copy Detection: a Comparative Study" CIVR (2007) Jul. 9-11 ACM pp. 1-8. *
Office Action dated Mar. 5, 2013 issued by the State Intellectual Property Office of the People's Republic of China in counterpart Chinese Patent Application No. 201080005948.6.
Solberg et al. "A Study of the Invariance Properties of TExtual Features in SAR Images." 1995 International Geoscience and Remote Sensing Symposium, IGARSS '95. Quantitative Remote Sensing for Science and Applications (Cat. No. 95CH35770) IEEE New York, NY, vol. 1, pp. 670-672, 1995.

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120014609A1 (en) * 2009-03-13 2012-01-19 Nec Corporation Image signature extraction device
US8744193B2 (en) * 2009-03-13 2014-06-03 Nec Corporation Image signature extraction device
US20160042517A1 (en) * 2014-08-08 2016-02-11 Robotic Vision Technologies, LLC Detection and tracking of item features
US9734401B2 (en) * 2014-08-08 2017-08-15 Roboticvisiontech, Inc. Detection and tracking of item features

Also Published As

Publication number Publication date
KR101404401B1 (ko) 2014-06-05
JPWO2010087124A1 (ja) 2012-08-02
EP2333718B1 (fr) 2013-08-28
EP2333718A1 (fr) 2011-06-15
WO2010087124A1 (fr) 2010-08-05
KR20110103423A (ko) 2011-09-20
JP4766197B2 (ja) 2011-09-07
EP2333718A4 (fr) 2011-11-09
CN102301395A (zh) 2011-12-28
CN102301395B (zh) 2014-08-06
US20110135203A1 (en) 2011-06-09

Similar Documents

Publication Publication Date Title
US8620087B2 (en) Feature selection device
Kabbai et al. Image classification by combining local and global features
Alzate et al. Multiway spectral clustering with out-of-sample extensions through weighted kernel PCA
Pietikäinen et al. Two decades of local binary patterns: A survey
Wang et al. Exploring DCT coefficient quantization effects for local tampering detection
Chen et al. Using binarization and hashing for efficient SIFT matching
Taimori et al. A novel forensic image analysis tool for discovering double JPEG compression clues
O’Connor et al. Facial recognition using modified local binary pattern and random forest
Li et al. Robust image hashing based on low-rank and sparse decomposition
US9747521B2 (en) Frequency domain interest point descriptor
US8170341B2 (en) Image signature extraction device
Roy et al. Digital image forensics
Grana et al. Layout analysis and content enrichment of digitized books
Tang et al. Robust image hashing via visual attention model and ring partition
Khodadoust et al. Partial fingerprint identification for large databases
Jouili et al. Graph embedding using constant shift embedding
WO2007026951A1 (fr) Procédé et dispositif de recherche d’image
Dinu et al. Local patch dissimilarity for images
Kashyap A novel method for real-time object-based copy-move tampering localization in videos using fine-tuned YOLO V8
Parnak et al. A Novel Image Splicing Detection Algorithm Based on Generalized and Traditional Benford’s Law
CN104700098B (zh) 一种基于改进型费希尔脸识别的暗环境人脸识别方法
Voloshynovskiy et al. On accuracy, robustness, and security of bag-of-word search systems
Bhagyalakshmi et al. Image retrieval using color and texture binary patterns
Lee et al. An identification framework for print-scan books in a large database
Riesen et al. Reducing the dimensionality of vector space embeddings of graphs

Legal Events

Date Code Title Description
AS Assignment

Owner name: NEC CORPORATION, JAPAN

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:IWAMOTO, KOTA;OAMI, RYOMA;REEL/FRAME:025724/0424

Effective date: 20110106

AS Assignment

Owner name: NEC CORPORATION, JAPAN

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:IWAMOTO, KOTA;OAMI, RYOMA;REEL/FRAME:025751/0066

Effective date: 20110106

STCF Information on status: patent grant

Free format text: PATENTED CASE

CC Certificate of correction
FPAY Fee payment

Year of fee payment: 4

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 8TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1552); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Year of fee payment: 8